Hi,
On Thu, Oct 18, 2007, Nicolas Boullis wrote:
> Do you really think it's worth bothering the release team for such a
> small library transition (it affects something like 5 source packages)?
Yes, I do. It affects 8 packages (directly).
> I thought I'd rather ask the individual maintainers to re-upload that
> bother the release team.
I beg to differ; bugging 6 packages, 6 different maintainers (I
excluded vcdimager), and making 6 stupid manual builds + source uploads
is way more time consuming than scheduling a bin NMU at a single place.
Also, you shouldn't do a sourceful upload when the source doesn't
change.
I also believe that it's a good idea to talk to the release team for
other reasons than just the rebuilds: coordinating the transition is in
general a good idea; less than 5 source packages perhaps doesn't need
coordination effort (but warning them still makes sense IMO), but it's
already 8 sources in total that will need to migrate together. There
are other implications than simply these 8 source packages: this
transition might get entangled into another one by any of the affected
source packages or their rdeps or rbdeps.
> Moreover, the maintainers would the have the
> opportunity to check that the transition does not break anything. I
> thought I'd wait a few days, and then file bugs against affected
> packages, asking for a re-upload.
Testing can only happen after a rebuild; I think proper testing already
happens when the rebuild is available for everybody to test. The way
you're handling this transition is that your libcdio package aged 9
days, then you might start filing bugs, then maintainers start doing
uploads, then their packages age 10 days, then perhaps if nothing has
popped up everything moves to testing. I believe this manner of doing
the transition inflicts a minimum of one month wait (and that's a
conservative minimum) and effectively delays testing: not only are
rebuilt packages available later, but they will also reach our
"testing" distribution in a long time, hence augmenting the risk of
being entangled by another transition.
For example, are you aware that gst-plugins-good0.10 entangles your
transition to gtk+2.0 2.12 which wont reach testing until some weeks
still? The release would know this.
Also, you didn't consider uploading to experimental first and
introduced a completely random delay in the transition which means your
libcdio upload might have been entangled with /anything/ in unstable
when it was accepted.
> If you disagree and would not accept to re-upload, feel free to bother
> the release team yourself.
I "bothered" the release team myself, no thanks for putting the burden
of your transition on my shoulders.
Bye,
--
Loïc Minier
